The 13 biggest announcements at Google I/O 2026

Google’s I/O 2026 keynote today was once again full of AI-related announcements including a new family of Gemini 3.5 AI models, new features for Search and Gmail, and updates about its Project Aura smart glasses.

The 13 biggest announcements at Google I/O 2026

Another AI-heavy Google keynote featured updates across its apps and AI models.

Another AI-heavy Google keynote featured updates across its apps and AI models.

If you weren’t able to tune into the event’s livestream today or follow along with our live blog, you can catch up on everything you missed in our roundup below.

Google launched updated AI models at I/O, starting with Gemini 3.5 Flash, with Gemini 3.5 Pro following next month. Starting today, Gemini 3.5 Flash will be the default model for the Gemini app and AI Mode in Search. Google says the new model is significantly faster, better at handling agentic tasks, offers improved agentic coding capabilities, and generates “richer, more interactive web UIs and graphics.” It also has improved guardrails that, according to Google, will make Gemini 3.5 Flash less likely to generate harmful content and less likely to accidentally flag safe queries as unsafe.

Along with Gemini 3.5 Flash, Google is also launching a redesign for the Gemini app. The new look, which Google calls “neural expressive,” features new animations, pops of color, a new font, and haptic feedback. It’s rolling out starting May 19th on the web and in the Gemini app on Android and iOS.

Along with Gemini 3.5, Google is launching an entirely new family of AI models it calls Gemini Omni. The first, Omni Flash, is rolling out starting today in the Gemini app, Google Flow, and YouTube Shorts. At launch, Omni Flash will be able to generate video clips from prompts that include a variety of inputs including text, photos, video, and audio (unlike Google’s Veo model, which is only text to video). Down the line, Google says Omni will be able to “create anything from any input.”

The new always-on AI agent that can write emails, create study guides, and keep an eye out for hidden credit card fees for you is Google’s “we have OpenClaw at home.” Powered by Gemini 3.5 Flash, Gemini Spark runs in the background using virtual machines on Google Cloud 24/7 and can connect to Google Workspace apps including Docs, Gmail, Sheets, and Slides, as well as third-party apps like Canva and Instacart. Google also plans to expand Spark’s capabilities with access to local files through the Gemini app on macOS.

Google will let users vibe-code entire native Android apps now, which they can eventually even publish to the Play Store straight from AI Studio. The feature includes an embedded Android emulator for viewing and editing app previews. Users can also plug in their phone to install their app directly from AI Studio for testing. Apps can be exported to Android Studio or GitHub, or saved as a ZIP file.

Google says it will also “soon” allow users to publish vibe-coded apps exclusively for friends and family, rather than publicly. Support for Firebase integrations is also coming later.

Google showed off an updated version of its Project Aura smart glasses, which it’s collaborating on with Xreal. The glasses were originally unveiled last year, but The Verge’s Victoria Song got a look at an updated version at I/O. The external compute puck for the glasses has been redesigned, and a fingerprint sensor has been added to it (along with a lanyard so you can wear the puck and the glasses). We also got to try out some new features in Google’s XR platform, including widgets for display glasses, Gemini integrations with Google Calendar and Google Keep, and improved Gemini performance.

Along with Project Aura, two new pairs of Android XR smart glasses are launching this fall, one from Warby Parker and one from Gentle Monster. Google announced partnerships with both brands at last year’s I/O, and now we finally know what those glasses will look like. Similar to the base Ray-Ban Meta smart glasses, these will be audio-only (no display). They’ll support features like live translation, navigation assistance with Gemini, and notification summaries.

Google is launching a “Universal Cart” that users can add products to from YouTube, Search, Gemini, and Gmail. Google says this “intelligent shopping cart” works across different merchants and services, like Nike, Target, Walmart, Ulta Beauty, Sephora, Wayfair, and Shopify. So, you could add products from Nike and Target to your Universal Cart and check out from both at the same time.

According to Google, Universal Cart can spot potential issues with products in your cart, as well, like flagging incompatible parts for a gaming PC. It can also interpret perks and loyalty info from Google Wallet payment methods to help users find savings or earn points. Universal Cart is launching in Search and Gemini this summer, with YouTube and Gmail coming later.

Google is expanding Gmail’s already robust search tools with a version of the Gemini Live experience for your inbox. Clicking an icon in the search bar lets you ask questions by talking, and instead of having to wade through a list of relevant email chains, the Gmail Live interface will extract and deliver pertinent information based on your prompt. It could be a faster way to pull up a specific confirmation code for a hotel reservation, for example. Similar voice-driven AI features will also be coming to Google Docs and Keep and be able to pull in data from Google Drive and Gmail.

Google Workspace is getting a new app called Pics designed to make it easier to apply iterative AI-generated updates to images. Pics is powered by Nano Banana 2 and Gemini and allows users to make changes by clicking on a specific part of an image and leaving a comment describing the edits, instead of having to write an entire prompt that includes the required updates. Eventually, Google plans to incorporate Pics’ capabilities into other Workspace apps to further streamline image editing.

Google is changing up the search box. Now it will expand to give users more space for longer queries and include AI-generated suggestions, similar to autocomplete. Users aren’t limited to searching with text, either — Google will now let you use text, images, files, videos, and even Chrome tabs as search input.

Search is also getting “information agents,” which can give you summarized updates on specific topics or questions, pulling from blogs, news, social media posts, and more. These agents will be launching this summer, starting with AI Pro and Ultra subscribers.

Google’s new “generative UI” feature can generate visuals in Search, like simulations or interactive tables and graphs. Search will also soon be able to generate “mini apps” for tasks or topics users search for multiple times, like a custom dashboard for planning an event.

When it was introduced last year at I/O 2025, Google’s premium AI Ultra subscription cost $249.99 per month for access to its most advanced AI models, higher usage limits, and early access to new tools. Google is now introducing two new pricing tiers for AI Ultra to match OpenAI. It now starts at $100 per month, while a $200 per month option includes access to Google’s Project Genie.

In an attempt to make it easier for people to identify and confirm when images have been generated or altered by AI, Google is expanding its AI detection tools to Chrome and Search. Starting today, uploading or selecting online images with Search (including tools like Google Lens and Circle to Search) can reveal more details about how or where the image was created using Google’s SynthID watermarking technology and C2PA Content Credentials. Additional functionality through Chrome that lets you circle questionable images on websites to see their provenance is coming later.

After renaming Project Starline to Google Beam last year, Google is now experimenting with lifelike AI agents that can do more than just talk to you during a call. The Verge’s Sean Hollister got to try an early demo of Sophie, a Beam video agent designed to make interactions with AI feel more personable. Although Sophie didn’t appear in 3D like when you’re talking to another person using the $25,000 HP Dimension system, the agent can respond to questions, read documents held up to the camera, and look up things like restaurant recommendations. During our demo, Google also demonstrated group calls using Beam — a feature that’s been in development for a few years and will work with tools like Google Meet and Zoom.

Follow topics and authorsfrom this story to see more like this in your personalized homepage feed and to receive email updates.